Detection Of Driver Fatigue Based On Biological Signals

Driver fatigue is one of the important causes of serious traffic accidents.Therefore, the research in detection methods of driving fatigue has become animportant issue.Firstly, we introduce visual and auditory stimuli tasks and design the drivingsimulator experiment with8stages to study detection of driver fatigue based onbiological signals. During the experiment,19subjects were asked to perform drivingabout two hours, and simultaneous acquisition of EEG and ECG. After the end ofeach stage, the subjects were told to finish Karolinska sleepiness scale andNASA-TLX scale as a reference for the result of signal analysis. Then biologicalsignals from driving simulator are studied from the following four aspects in thispaper:1) Eye blink is an important and inevitable artifact during scalp EEG recording.So we proposed a method based on correlation index and the feature of powerdistribution to automatically detect eye blink components. Furthermore, thecorrelation between independent components and scalp EEG channels can betranslating directly from the mixing matrix of ICA. This helps to simplify calculations.The proposed method doesn’t need to select a template or threshold in advance, and itworks without simultaneously recording an electrooculography (EOG) reference. Theexperimental results demonstrate that the proposed method can automaticallyrecognize eye blink components with a high accuracy.2) Study the relationship between driver fatigue and power spectrum of the EEGband. The physiological indicators of19subjects before and after driving, were takenfor statistical significance test, the results of the experiment showed that the EEGspectral indicators can effectively detect driver fatigue and alert.3) This paper applied the concept of phase synchronization to investigate EEGfeatures in driver fatigue. Mean phase coherence (MPC) is employed as a measure forphase synchronization. The analysis mainly focused on finding the spatial-frequencyfeatures associated with mental state shifted. The major finding is that the phasesynchronization of delta and alpha bands significantly enhanced in the frontal andparietal lobes. The statistical analysis results suggest MPC can significantlydistinguish different mental states such as alert and fatigue. In addition, the results of experiment also demonstrate that a simple spatial-frequency pairs of electrodes, i.e.,Pz-Fz in delta band, helps to expand the real-world application of EEG in theevaluation of driver fatigue.4) In ECG data analysis, heart rate variability (HRV) indexes in the time domainand frequency domain were studied and discussed. After detecting R-wave usingdifferential threshold method, RR interval sequence was obtained， then various HRVindexes were studied. With a deeper level of fatigue, some HRV indexes in the timeand frequency domain changed significantly. The results provide another importantreference for the use of ECG in the detection of driver fatigue.At last, on the basis of previous signal analysis, the fusion of EEG and ECGcharacteristics is discusses. Studied the data fusion and classification method based onsupport vector machine, and achieved fusion of EEG and ECG in the feature level.Experimental results show that the EEG and ECG fusion can be effectively improvedrecognition rate in classification.